We work on a variety of topics related the perception of natural stimuli, which we study with electro-encephalography (EEG). We are also trying to understand basic mechanisms of transcranial electric stimulation (TES) on learning, as well as develop tools for targeting TES. We also have projects on deep-learning to analyze MRI images.

Neural processing of natural stimuli

...

We are interested in how people process natural stimuli and use EEG to measure stimulus-evoked neural activity. We sometimes collect also eye movements, pupil and the heart beat as they are remarkably connected to what we do in our minds. In our experience, small changes in the task can have large effects on the resulting brain signals. We therefore emphasize the most natural setting possible. Video and audio narratives provide a good balance between reproducibility and natural dynamics. We have also started to look at video games and music. The workhorse for the analysis of these natural stimuli has been inter-subject correlation (ISC) which does not require any kind of labeling of the stimulus.

Current projects:

  • Attention in online educational videos: Jens Madsen
  • Attention to speech: Ivan Iotzov
  • Synchronization of physiological and behavioral signals: Jens Madsen
  • Synchronization of neural acivity to music: Jens Madsen
Engagement with video | Cohen et al. 2017 | Dmochowski et al. 2014 | Dmochowski et al. 2012

ISC of EEG appears to be a marker of engagement with a video stimulus. It is therefore predictive of the behavior of large audiences, including tweeting, viewership size and preference ratings. Engaging narratives synchronize not just our brains, but also our perception of time.

Attention and memory for narratives | Ki et al. 2016 | Cohen et al. 2016 | Cohen and Madsen et al. 2018

ISC of EEG is dramatically modulated by attention.

Video games | Dmochowski et al. 2016

For the case of unique experience we find that correlation with the actual stimulus also captures attention. Surprisingly, we find a strong coupling of the stimulus with a supramodal component of the EEG.

Videos in the classroom | Poulsen et al. 2016

ISC of EEG during videos can be measured simultansously in the classroom.

Synchronized eye movements | Burleson-Lesser et al. 2017 | Madsen et al. 2021

Synchronized eye movements are modulated by attention and can predict test scores of student in online video education. Attention can be measured remotely using standard webcameras without the need to transmit any personal information, thus preserving privacy. See video exampes

Synchronized heart rate | Perez and Madsen et al. 2020

Narratives synchronize heart rate between individuals and could be a predictor of consciousness.

Engagement with music | Madsen et al. 2019

ISC of EEG to people listening to music is modulated by attention and is influenced by peoples knowledge about the music and their musical training. See video exampes

Mechanisms of TES

...

Transcranial electrical stimulation (TES) applies weak electric currents at the scalp in order to improve brain function. While this has shown a great deal of promise in clinical and cognitive studies,there is very limited knowledge of the underlying cellular mechanisms. We aim to understand how stimulation modifies information processing and storage in individual neurons and networks. In the past we have demonstrated that stimulation acutely modifies firing rate, spike timing, excitability, synaptic efficacy, and network oscillations. More recently we have focused on how these acute effects translate into long-term strorage through synaptic plasticity.

Current projects:

  • Modulation of hippocampal synaptic plasticity with direct current stimulation: Greg Kronberg
Acute effects of tDCS | Reato et al. 2010 | Rahman et al. 2013 | Lafon et al. 2016 Acute effects of tACS | Reato et al. 2013

Long-term effects of tDCS | Reato et al. 2015 | Kronberg et al. 2017

Targeting of TES with electrode arrays

... ... ...

TES is applied often using simple sponge electrodes. We have advocated the use "high-definition" stimulation using arrays of small electrodes (comparable to what is done with EEG). To this end we have developed methods to steer currents with these arrays, as well as build individualized head models in particular to account for altered anatomies in stroke patients. We have been the first to thoroughly validate the corresponding current-flow models in human.

Current projects:

Automated whole-head segmentation | Huang et al. 2013 | Huang, Parra 2015

Model validation | Huang, Liu, et al. 2017

TES targeting | Dmochowski et al. 2011 | Dmochowski et al. 2016

Deep-learning for MRI

...
...

We developed deep-learning architectures to analyze 3D images, specifically for brain segmentation, and breast tumor segmentation. Head and brain segmentation is of particular interest because of the current flow-modeling describe above. A collaboration with Memorial Sloan Katering Cancer Center involves projects related to breast and brain MRI.

Current projects:

  • Automated brain segmentation in the presence of lesions
  • Breast cancer segmentation
Head segmentation | Hirsch et al 2019

Breast cancer | Hirsch et al 2020